An Introduction to Frequent Pattern Mining - Frequent Pattern Mining - page 13

6 Applications of Frequent Pattern Mining

Frequent pattern mining has applications of two types. The first type of application
is to other major data mining problems such as clustering, outlier detection, and
classification. Frequent patterns are often used to determine relevant clusters from
the underlying data. In addition, rule-based classifiers are often constructed with
the use of frequent pattern mining methods. Frequent pattern mining is also used
in generic applications, such as Web log analytics, software bug analysis, chemical,
and biological data.

6.1 Applications to Major Data Mining Problems

Frequent pattern mining methods can also be applied to other major data mining
problems such as clustering [ 9 , 19 ], classification and outlier analysis. For example,
frequent pattern mining methods are often used for subspace clustering [ 11 ], by
discretizing the quantitative attributes, and then finding patterns from these discrete
values. Each such pattern, therefore, corresponds to a rectangular region in a subspace
of the data. These rectangular regions can then be integrated together in order to create
a more comprehensive subspace representation.
Frequent pattern mining is also applied to problems such as classification, in
which rules are generated by using patterns on the left hand side of the rule, and
the class variable on the right hand side of the rule [ 52 ]. The main goal here is
to find discriminative patterns for the purpose of classification, rather than simply
patterns that satisfy the support requirements. Such methods have also been extended
to structured XML data [ 73 ] by finding discriminative graph-structured patterns. In
addition, sequential pattern mining methods can be applied to other temporal mining
methods such as event detection [ 43 , 44 , 53 , 54 ] and sequence classification [ 68 ].
Frequent pattern mining has also been applied to the problem of outlier analysis
[ 1 ], by determining deviations from the expected patterns in the underlying data.
Methods for clustering based on frequent pattern mining are discussed in Chap. 16,
while rule-based classification are discussed in Chap. 17. It should be pointed out that
constrained frequent pattern mining is closely related to the problem of classification
with frequent patterns, and therefore both are discussed in the same chapter.

6.2 Generic Applications

Frequent pattern mining has applications to a variety of problems such as clustering,
classification and event detection. In addition, specific application areas such as Web
mining and software bug detection can also benefit from frequent pattern mining
methods. In the context of Web mining, numerous methods have been proposed for
finding useful patterns from Web logs in order to make recommendations [ 63 ]. Such